Skip to main content

Amine AWADA

Thursday, 19 of Decembe, 2024

Activities co-definition in a smart environment: human-system perception alignment

ABSTRACT
Energy consumption has increased drastically in the last decades. Our everyday life becomes more and more dependent on energy-consuming machines and devices. Studies in this field, show that buildings are an essential concern because they consume a lot of energy. Even though, new regulations and laws made buildings more efficient energetically speaking, the consumption remains high, which indicates that it is affected by another factor, the human factor (the occupant). The occupants' behavior can be inappropriate energetically speaking and can lead to huge amounts of waste in energy consumption, so it is necessary to detect these behaviors in order to teach and notify the occupants about the inefficient ones and suggest alternative ways (if existing) to better achieve their activities without spending a lot of energy. Many applications and researches were made to integrate the occupants in the energy saving process. For example generating explanations for the occupants about their energy use and how they can alter their actions in order to reduce their consumption while maintaining a good comfort level. These applications and researches require a good knowledge on human activities. Human activity recognition is an important field of research, its importance is not limited to the energy field but transcends to healthcare, home security and other fields of study. Thanks to modern smart home technologies, collecting data about occupants to infer their activities, became easier and socially acceptable. Still, activities' labels collection is somehow overwhelming in most cases requiring lots of efforts and time. Adding that knowledge (data, labels and correlation functions between them) transfer from system to another is not practical since labels varies from occupants to others and, buildings architecture differences affect installed sensors data. Taking into consideration the differences in perception between human beings and machines, the objective in this thesis project is to build a dynamic system capable of efficiently recognizing human activities using ambient sensors installed in a smart environment based on a co-definition approach (a cooperation process between the human agent and artificial system) to align its perception with the human's perception. The system should also be capable of dealing with different kinds of human (single or multi occupants, forgetting and mistakes in labeling, labels differences according to occupants' contexts) and equipment (different buildings architecture, sensors' numbers and positions) related problems.

Date and place

Thursday, 19 Décember at 14:00

Grenoble INP, Amphithéâtre Gosse
and visio

Jury members

Pierre DE LOOR
PROFESSEUR DES UNIVERSITES, École Nationale d'Ingénieurs de Brest Rapporteur
Mora LAURENT
PROFESSEUR DES UNIVERSITES, IUT de Bordeaux Rapporteur
Maria DI MASCOLO
DIRECTRICE DE RECHERCHE, CNRS Délégation Alpes Examinatrice
Yann LAURILLAU
PROFESSEUR DES UNIVERSITES, Université Grenoble-Alpes / IUT2 Examinateur
Patrick REIGNIER
PROFESSEUR DES UNIVERSITES, GRENOBLE INP Directeur de thèse
Stéphane PLOIX
PROFESSEUR DES UNIVERSITES, GRENOBLE INP Co-directeur de thèse
EL Abed EL SAFADI
Co-encadrant de thèse

Submitted on December 16, 2024

Updated on December 16, 2024